Logistic Regression Model Fit Statistics

1. Omnibus Test of Model Coefficients

• It uses the likelihood ratio chi-square (LR chi²) statistic:

  χ² = -2LL(null) - (-2LL(model)) = -2 × [LL(null) - LL(model)]

• Degrees of freedom = number of predictors in the model.
• A significant p-value (< 0.05) means the predictors, taken together, significantly improve the model's predictive ability over chance.

Think of this as the overall F-test equivalent from linear regression, but using chi-square instead.

2. Nagelkerke R²

1. Cox & Snell R² is the base version, computed as:

   R²_CS = 1 - [L(null) / L(model)]^(2/n)

   Its maximum value is less than 1 (capped depending on the data), making it hard to interpret on a 0-1 scale.
2. Nagelkerke R² corrects this by dividing Cox & Snell R² by its theoretical maximum:

   R²_N = R²_CS / R²_max, where R²_max = 1 - L(null)^(2/n)

   This forces the range to 0 to 1, making it more interpretable.

3. Hosmer-Lemeshow Test

1. Cases are ranked by predicted probability and split into 10 equal-sized groups (deciles).
2. Within each group, observed vs. expected event counts are compared using a chi-square-like statistic.
3. df = number of groups - 2 = 8 (typically).

H-L χ² = Σ [(O_k - E_k)² / E_k(1 - E_k/n_k)]

• Sensitive to sample size: very large samples may produce significant p-values even for trivially small deviations
• Results can differ based on the number of groups chosen
• It tests calibration only - a well-calibrated model can still discriminate poorly (complement with AUC/ROC)

How They Work Together

• Omnibus χ² = 23.4, p < 0.001 (predictors are significant)
• Nagelkerke R² = 0.17 (model explains only 17% of variance - small effect)
• H-L χ² = 7.3, df = 8, p = 0.50 (good calibration - predictions match observations)

Quick Reference

Omnibus test  →  "Is the model better than nothing?"         (overall significance)
Nagelkerke R² →  "How much does the model explain?"          (effect size)
H-L test      →  "Are the predicted probabilities accurate?" (calibration)

Tuberculosis and Diabetes - Collaborative Programmes

1. The TB-Diabetes Syndemic: Why It Matters

• Diabetes accounts for ~20% of all TB cases and ~10% of smear-positive TB globally
• People with diabetes have a 2-3x higher risk of progressing from latent to active TB compared to non-diabetics (Park's Textbook of Preventive & Social Medicine)
• The 2024 WHO Global TB Report attributes ~400,000 TB episodes annually to diabetes worldwide
• TB worsens glycemic control, and diabetes worsens TB outcomes - creating a vicious cycle

Mechanistic link:

2. WHO Collaborative Framework

"People with diabetes face a heightened risk of developing TB and suffering from adverse treatment outcomes. Ensuring universal access to screening, prevention and care for both TB and diabetes is essential."

• Dr Tereza Kasaeva, WHO Global TB Programme Director, 2025

3. National Framework for Joint TB-Diabetes Collaborative Activities

A. Improve Diagnosis and Management of Diabetes Among TB Patients

• Screening all registered TB patients for diabetes mellitus - using fasting blood glucose (FBG) or HbA1c at TB diagnosis
• Ensuring diabetes management is integrated into TB treatment (glycemic control throughout TB therapy)

B. Improve Diagnosis and Management of TB Among Diabetic Patients

• Intensified detection of active TB disease among patients attending diabetes clinics
• Ensuring TB infection control measures in healthcare settings where diabetes is managed
• Ensuring proper TB treatment and follow-up in comorbid patients

C. Joint Monitoring and Evaluation

• Shared registers and reporting systems to track TB-DM co-morbid patients
• Monitoring bidirectional screening coverage and outcomes
• Evaluating treatment success rates in co-morbid patients vs. TB-only patients

D. Treatment Considerations

• TB patients diagnosed with diabetes should receive the same duration of TB treatment with daily regimen as non-diabetic patients (no modification to treatment length is currently recommended based on this alone)

4. Bidirectional Screening: The Core Strategy

5. Impact of Diabetes on TB Outcomes

6. Impact of TB on Diabetes

• TB illness causes physiological stress that raises blood glucose (counter-regulatory hormones)
• Anti-TB drugs, particularly rifampicin, induce hepatic enzymes that reduce levels of sulfonylureas and other oral hypoglycemics - dosage adjustments may be needed
• Steroid treatment (used in TB meningitis/pericarditis) worsens glycemic control
• Weight loss from TB further destabilizes glucose metabolism

7. Programme Integration - Key Challenges

1. Vertical programmes - TB and diabetes are traditionally managed by separate departments with separate registers and staff
2. Resource constraints in high-burden, low-income settings
3. Cost evidence gaps - the BMC 2024 review noted insufficient economic data
4. Inconsistent screening coverage - many countries screen TB→DM but not DM→TB
5. Stigma attached to both conditions, leading to delayed care-seeking

8. The India Context (Park's)

Summary

1. Bidirectional screening at both TB clinics and diabetes clinics
2. Co-management with shared monitoring
3. No change in TB treatment duration for diabetics, but close glycemic monitoring
4. Integration of vertical programmes for health system strengthening

• Park's Textbook of Preventive and Social Medicine, p. 236
• Franco et al., Cochrane Database Syst Rev 2024 - PMID 39177079
• Milice et al., BMC Public Health 2024 - TB-T2D collaborative framework review
• WHO Operational Handbook on TB and Comorbidities: Diabetes, Feb 2025

National Programme for Prevention and Control of Diabetes (India)

Programme Evolution

Rationale

• NCDs account for ~63% of all deaths in India
• Diabetes alone accounts for ~3% of total mortality
• India has one of the world's largest diabetes burdens (~100+ million people with diabetes)
• The existing health system was historically focused on communicable diseases - NPCDCS was designed to reorient it toward NCDs

Major Objectives

1. Prevent and control common NCDs through behaviour and lifestyle changes
2. Provide early diagnosis and management of common NCDs
3. Build capacity at various levels of healthcare for prevention, diagnosis and treatment
4. Train human resources - doctors, paramedics and nursing staff
5. Establish capacity for palliative and rehabilitative care

Programme Structure - Tiered Delivery

Sub-Centre Level

• Health promotion - camps, interpersonal communication, posters, banners
• Opportunistic screening of all persons above 30 years:
  • BP measurement
  • Blood glucose by strip method (glucometer)
• Suspected cases of diabetes and hypertension referred to CHC/higher facility

Community Health Centre (CHC) Level

• NCD Clinic established at each CHC
• Confirmed diagnosis and management of diabetes and hypertension (outpatient and inpatient)
• Nurse appointed under programme conducts home visits for bedridden cases
• Supervises health workers
• Complicated cases referred to District Hospital

District Hospital Level

• NCD Clinic - screens all persons >30 years for diabetes, hypertension, CVD
• Detailed investigations for high-risk individuals and referrals from CHC
• Regular management and annual assessment of confirmed cases
• Management of established cardiovascular disease
• Home-based palliative care for chronic/debilitating cases
• Health education and counselling for patients and attendants

Key Strategies

Urban Health Check-up Scheme for Diabetes and Hypertension

1. Screen urban slum population for diabetes and high blood pressure
2. Create a database of prevalence of DM and hypertension in urban slums
3. Sensitize urban slum population about healthy lifestyle

Implementation Infrastructure

Programme Progress (as of 2015):

NP-NCD 2023-2030 (Updated Programme)

• Retains all NPCDCS components (diabetes, hypertension, cancer, CVD, stroke)
• Adds CKD and NAFLD as priority conditions
• Strengthens integration with Health and Wellness Centres (HWCs) under Ayushman Bharat
• Population-based screening continues for persons >30 years for diabetes, hypertension, and common cancers (oral, breast, cervical)
• Delivered through trained community health officers at HWCs

NCD Cell Structure

• National NCD Cell - policy, funding, monitoring
• State NCD Cell - state-level implementation
• District NCD Cell - district-level coordination and reporting

Integration with Other Programmes

Key Messages for Behaviour Change Communication

• Increased intake of healthy foods (reduce salt, sugar, saturated fat)
• Increased physical activity (at least 30 min/day)
• Avoidance of tobacco and alcohol
• Stress management

Diabetes Diagnostic Criteria and Targets - India's NP-NCD Programme

1. Diagnostic Criteria

Classification by Glucose Values

Note: Random plasma glucose ≥200 mg/dL is diagnostic only when accompanied by classic diabetes symptoms (polyuria, polydipsia, unexplained weight loss). Otherwise, a confirmatory test is needed.

Screening Tool Used at Field Level

2. Special Criteria - Pregnancy

• All pregnant women are screened in the 1st trimester with FPG
• GDM threshold is deliberately lower (≥92 mg/dL) than the non-pregnant diagnostic cut-off

3. Risk Factors Triggering Screening (NP-NCD Programme)

• Overweight/obese (BMI > 25)
• Recurrent infections (tinea, oral thrush, UTI, cellulitis, carbuncle)
• Non-healing ulcers (foot ulcers - neuropathic/infected)
• Exogenous/iatrogenic Cushing's syndrome
• History of GDM or pre-existing diabetes in pregnancy
• Recurrent sinusitis, STIs, osteomyelitis, septic arthritis

4. Treatment Targets (Goals of Control)

5. Monitoring Schedule

6. Pharmacotherapy Algorithm (HbA1c-based)

First-line Drug - Metformin (Biguanide):

• Daily dose range: 250-2500 mg (1-3 times/day)
• Metformin SR: 500-2500 mg (1-2 times/day)
• Contraindications: Renal insufficiency, hepatic insufficiency, respiratory insufficiency, acute MI, congestive cardiac failure, alcohol abuse, ketoacidosis, severe infections

Once diabetes is diagnosed, the patient must be provided at least one month's supply of drugs at the time of diagnosis.

Summary Table

DIAGNOSIS         FASTING     2-hr PPG    HbA1c    RANDOM + SYMPTOMS
Normal            <100        <140        <5.7%    —
Pre-diabetes      100-125     140-199     5.7-6.4% —
Diabetes          ≥126        ≥200        ≥6.5%    ≥200 mg/dL + symptoms

TARGETS
HbA1c             ≤7% (individualized)
FPG               80-130 mg/dL
2-hr PPG          <180 mg/dL
BP                <140/90 (or <130/80 if CKD)
LDL               <100 mg/dL (<70 if CAD)